Power grid assets prediction using generative adversarial networks

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using a neural network to predict locations of feeders in an electrical power grid. One of the methods includes training a generative adversarial network comprising a generator and a discriminator; and generating, by the generator, from input images, output images with feeder metadata that represents predicted locations of feeder assets, including receiving by the generator a first input image and generating by the generator a corresponding first output image with first feeder data that identifies one or more feeder assets and their respective locations, wherein the one or more feeder assets had not been identified in any input to the generator.

What is claimed is: 
     
       1. A method for using a neural network to predict locations of feeders in an electrical power grid, the method comprising:
 training a generative adversarial network comprising a generator and a discriminator, wherein training the generative adversarial network comprises:
 training the generator while holding the discriminator fixed, including providing training input to the generator, the training input comprising training input map data, and 
 providing corresponding training outputs generated by the generator to the discriminator, the training outputs comprising map data including feeder metadata, and training the generator based on a respective discriminator output from the discriminator for each training output received by the discriminator; and 
 
 generating, by the generator, from input map data including input images, output map data with feeder metadata that represents predicted locations of feeder assets, including receiving, by the generator, a first input image and generating by the generator corresponding first output map data with first feeder data that identifies one or more feeder assets and their respective locations, wherein the one or more feeder assets had not been identified in any input to the generator. 
 
     
     
       2. The method of  claim 1 , wherein generating first output map data with first feeder data comprises also receiving with the first input image corresponding input feeder metadata representing one or more feeder assets in the first input image. 
     
     
       3. The method of  claim 2 , wherein generating the first output map data with first feeder data comprises also receiving with the first input image one or more asset placement rules. 
     
     
       4. The method of  claim 3 , wherein:
 the respective feeder data is incorporated in a respective output image. 
 
     
     
       5. The method of  claim 3 , wherein:
 the respective feeder data is generated as metadata separate from an output image. 
 
     
     
       6. The method of  claim 1 , wherein generating the first output map data includes generating, as first feeder data, data that identifies an underground feeder asset including a location of the underground feeder asset. 
     
     
       7. The method of  claim 6 , wherein the first feeder data identifies all feeder assets between a particular substation and a particular load. 
     
     
       8. The method of  claim 7 , wherein the particular load is a residential load. 
     
     
       9. The method of  claim 8 , wherein the one or more feeder assets comprise three or more of a line, a pole, a crossarm, a transformer, a switch, an insulator, a recloser, a sectionalizer, a capacitor bank, including switched capacitors, a load tap changer, or a tap. 
     
     
       10. The method of  claim 9 , wherein the one or more feeder assets comprise a first transformer and the first feeder data specifies a size of the first transformer. 
     
     
       11. The method of  claim 9 , wherein the one or more feeder assets comprise a first capacitor bank that includes switched capacitors. 
     
     
       12. The method of  claim 1 , wherein the generator and discriminator are each a respective convolutional neural network model. 
     
     
       13. The method of  claim 1 , wherein:
 the training input includes, for a first plurality of training inputs, data representing respective identified above-ground feeder assets corresponding to training input images. 
 
     
     
       14. The method of  claim 13 , wherein:
 the training input includes one or more asset placement rules. 
 
     
     
       15. The method of  claim 14 , wherein training the generative adversarial network comprises:
 training the discriminator while holding the generator fixed, including providing to the discriminator training output images generated by the generator and ground truth images, wherein the ground truth images provided to the discriminator each include respective ground truth feeder data identifying feeder assets and their locations on the ground truth images. 
 
     
     
       16. A system comprising one or more computers configured to perform operations comprising:
 training a generative adversarial network comprising a generator and a discriminator, wherein training the generative adversarial network comprises:
 training the generator while holding the discriminator fixed, including providing training input to the generator, the training input comprising training input map data, and 
 providing corresponding training outputs generated by the generator to the discriminator, the training outputs comprising map data including feeder metadata, and training the generator based on a respective discriminator output from the discriminator for each training output received by the discriminator; and 
 
 generating, by the generator, from input map data including input images, output map data with feeder metadata that represents predicted locations of feeder assets, including receiving, by the generator, a first input image and generating by the generator corresponding first output map data with first feeder data that identifies one or more feeder assets and their respective locations, wherein the one or more feeder assets had not been identified in any input to the generator. 
 
     
     
       17. The system of  claim 16 , wherein generating first output map data with first feeder data comprises also receiving with the first input image corresponding input feeder metadata representing one or more feeder assets in the first input image. 
     
     
       18. The system of  claim 17 , wherein generating the first output map data with first feeder data comprises also receiving with the first input image one or more asset placement rules. 
     
     
       19. The system of  claim 16 , wherein generating the first output map data includes generating, as first feeder data, data that identifies an underground feeder asset including a location of the underground feeder asset.